Discussion:
[Rabbit-dev] Rabbit-dev Digest, Vol 23, Issue 2
Rick Leir
2012-02-04 22:11:43 UTC
Permalink
I am a bit worried about the changes required
though, they would probably be quite big.
Some options in order of increasing complexity:

1/ replace client-side HTTP with SPDY, all users need chrome browser

2/ allow both protocols but not at the same time

3/ allow both protocols at the same time, but a client has to choose one
or the other

The Agile folks would tell us: Always start with the simplest possible
solution and get it working first. Option 1 does not look too difficult,
but you know best.

What features were you thinking of for Rabbit5?
cheers -- Rick
Robert Olofsson
2012-02-04 23:12:24 UTC
Permalink
On Sat, 4 Feb 2012 17:11:43 -0500
Post by Rick Leir
What features were you thinking of for Rabbit5?
Please note that there is no real plan and no work has started yet.

First thing I would like to do is to change how input and output
are handled. Currently we have a connection object that has in and out
for the client and in and out for the upstream server and the way this
is implemented means that full pipelining is hard to do. This will be a
major change and thus quite a rough ride. I think the code would end up
with a few, but simpler forwarders.

I would also like to update the cache so that we can cache both
original and modified content and thus easily can support client
settings for image compression. I am not sure how big this change
actually would be, not sure if it would be a 5.x issue, but I think it
would be.

I would also like to change some of the http header handling so that
we do not have to parse dates or cookies or ranges or ... multiple
times. Currently everything is stored as strings and that is not always
optimal, but makes some things easy.

/robo

Loading...